Background: High-dose short-term methylprednisolone is the recommended treatment in the management of multiple sclerosis relapses, although it has been suggested that lower doses may be equally effective. Also, glucocorticoids are associated with multiple and often dose-dependent adverse effects. This quantitative benefit-risk assessment compares high- and low-dose methylprednisolone (at least 2000 mg and less than 1000 mg, respectively, during at most 31 days) and a no treatment alternative, with the aim of determining which regimen, if any, is preferable in multiple sclerosis relapses. Methods: An overall framework of probabilistic decision analysis was applied, combining data from different sources. Effectiveness as well as risk of non-serious adverse effects were estimated from published clinical trials. However, as these trials recorded very few serious adverse effects, risk intervals for the latter were derived from individual case reports together with a range of plausible distributions. Probabilistic modelling driven by logically implied or clinically well motivated qualitative relations was used to derive utility distributions. Results: Low-dose methylprednisolone was not a supported option in this assessment; there was, however, only limited data available for this treatment alternative. High-dose methylprednisolone and the no treatment alternative interchanged as most preferred, contingent on the risk distributions applied for serious adverse effects, the assumed level of risk aversiveness in the patient population, and the relapse severity. Conclusions: The data presently available do not support a change of current treatment recommendations. There are strong incentives for further clinical research to reduce the uncertainty surrounding the effectiveness and the risks associated with methylprednisolone in multiple sclerosis relapses; this would enable better informed and more precise treatment recommendations in the future.
Sampled values for all probability variables of the benefit-risk assessment
The sampled values are based on probabilistic modelling of data from published clinical trials and from individual case safety reports contained in VigiBase, the WHO international database of suspected adverse drug reactions. The first two columns are used to determine what scenario of the assessment that is considered: the first column indicates the type of distribution over the derived risk intervals for the serious adverse effects, and the second column indicates the proportion of the sampled risk values that is attributed to the background. Each of the following columns belongs to one of the probability variables for one of the three considered treatment alternatives. Variables are listed in the order obtained when traversing the decision tree in Figure 3 top to bottom, starting from the leftmost level. There are 10,000 rows for each scenario. Abbreviations: HDMP=high-dose methylprednisolone, LDMP=low-dose methylprednisolone, NT=no treatment.
probabilities.csv
Sampled values for all utility variables of the benefit-risk assessment
These data have been generated by probabilistic modelling driven by qualitative relations specified between the considered clinical outcomes of the benefit-risk assessment. The first two columns are used to determine what scenario of the assessment that is considered: the first column indicates the severity of the relapse as measured by the patient's starting EDSS value, and the second column indicates the so called minimum utility difference between lethal and non-lethal outcomes. Each of the following columns belongs to the utility variable corresponding to one of the considered clinical outcomes. Variables are listed in the order obtained when viewing all branches of the decision tree in Figure 3 top to bottom. There are 10,000 rows for each scenario.
utilities.csv
Resulting samples for the expected utility of the three alternatives
The data displayed in this file are the resulting samples for the expected utility of the three considered treatment alternatives in all considered scenarios. They have been obtained by combining the data displayed in the two files 'Sampled values for all probability variables of the benefit-risk assessment' and 'Sampled values for all utility variables of the benefit-risk assessment', using the decision tree depicted in Figure 3 and the formulae given in Figure 1. The first four columns are used to determine what scenario of the assessment that is considered: the first column indicates the type of distribution over the derived risk intervals for the serious adverse effects; the second column indicates the proportion of the sampled risk values that is attributed to the background; the third column indicates the severity of the relapse as measured by the patient's starting EDSS value; and the fourth column indicates the so called minimum utility difference between lethal and non-lethal outcomes. The fifth, sixth, and seventh columns then show the computed expected utility values for high-dose methylprednisolone (HDMP), low-dose methylprednisolone (LDMP), and no treatment (NT), respectively. There are 10,000 rows for each scenario.
expected_utilities.csv
Resulting preference rates for the three alternatives
The data in this file has been computed directly from the raw data in the file 'Resulting samples for the expected utility of the three alternatives'. This file contains the same scenarios as that file, discernible from the the first four columns, but shows only a single value for each treatment alternative in each scenario. This value, the preference rate, is the fraction of all 10,000 sampling iterations for a given scenario in which the indicated alternative was the preferred one, i.e. had the highest expected utility. Preference rates for high-dose methylprednisolone (HDMP), low-dose methylprednisolone (LDMP), and no treatment (NT) are shown in columns 5-7, respectively. Figures 10 and 11 have been generated based on precisely this data.
preference_rates.csv
